Researchers have developed the first soft robot that can walk on rough surfaces such as sand and pebbles. The four-legged, 3-D printed bot's X shape layout allow it to have different types of walks for different terrains. The researchers say it could be used to record sensor data in dangerous environments or for search and rescue missions. The researcher tested the performance of the bot (pictured) with different leg configurations, gait sequences over various terrains, and it was able to successfully navigate over large rocks, under inclined surfaces and over small pebbles, walking at speeds up to 20 millimeters (0.8 inches) per second The robot, designed by researchers at the University of California in San Diego, was possible thanks to a high-end 3-D printer that allowed the researchers to print soft and rigid materials together within the same components. The robot's soft legs naturally conform to its surroundings during operation, resulting in a robot with the ability to crawl through a variety of terrains with limited computation and no sensing capabilities.

You have to jump before you can fly. A species of tiny parrot saves energy by hopping from branch to branch when foraging – a skill that may have helped bird ancestors to first get off the ground. These small birds hop between branches up to 30 times a minute, gaining propulsion from their legs and adding a few wingbeats to extend their range. A new study shows they do this in ways that minimise energy requirements, and suggests bird-like dinosaurs might have benefited from the technique too. To examine the biomechanics of these short flights, the Pacific parrotlets were trained to fly between perches for a food reward.

Who will remain a loyal customer and who won't? Which messages are most effective with which segments? How can customer value be maximized? This book supplies powerful tools for extracting the answers to these and other crucial business questions from the corporate databases where they lie buried. In the years since the first edition of this book, data mining has grown to become an indispensable tool of modern business.

Users will soon see Google's AI Assistant in a number of new devices -- including the Apple iPhone -- the company announced at its I/O conference in Mountain View, California on Wednesday. Assistant debuted in 2016 and was originally integrated into the company's Pixel phone and Home smart hub. It's since spread to more than 100 million individual devices, Google CEO Sundar Pichai told the crowd at this year's conference, including smart TVs, automobiles and wearables. And it's about to be in a lot more. While Assistant already works with more than 70 smart home device makers, the company announced that it will soon release an Assistant SDK so that developers hardware manufacturers can integrate the service into even more devices -- anything from speakers to blenders, smart locks to web cameras.

The concept of ensemble learning offers a promising avenue in learning from data streams under complex environments because it addresses the bias and variance dilemma better than its single model counterpart and features a reconfigurable structure, which is well suited to the given context. While various extensions of ensemble learning for mining non-stationary data streams can be found in the literature, most of them are crafted under a static base classifier and revisits preceding samples in the sliding window for a retraining step. This feature causes computationally prohibitive complexity and is not flexible enough to cope with rapidly changing environments. Their complexities are often demanding because it involves a large collection of offline classifiers due to the absence of structural complexities reduction mechanisms and lack of an online feature selection mechanism. A novel evolving ensemble classifier, namely Parsimonious Ensemble pENsemble, is proposed in this paper. pENsemble differs from existing architectures in the fact that it is built upon an evolving classifier from data streams, termed Parsimonious Classifier pClass. pENsemble is equipped by an ensemble pruning mechanism, which estimates a localized generalization error of a base classifier. A dynamic online feature selection scenario is integrated into the pENsemble. This method allows for dynamic selection and deselection of input features on the fly. pENsemble adopts a dynamic ensemble structure to output a final classification decision where it features a novel drift detection scenario to grow the ensemble structure. The efficacy of the pENsemble has been numerically demonstrated through rigorous numerical studies with dynamic and evolving data streams where it delivers the most encouraging performance in attaining a tradeoff between accuracy and complexity.

The ability of a footballer to outwit the goalkeeper depends in part on his ability to deliver the ball at an unpredictable time and location. Researchers have studied how we make sure such decisions are unpredictable, and found that the brain processes predictable and unpredictable components in different regions of the brain. This process ensures that we learn from experience, while still remaining spontaneous to get the competitive edge. Researchers set out to understand how the brain optimises the timing of actions to circumstance while retaining unpredictability. Readings were taken either in a region of the prefrontal cortex called MPFC, which is involved in decision-making learning, or in a region of the motor cortex, M2, thought to be involved in the direct control of movements.

In the Center for Clinical Data Science, teams from both companies will develop, test and deploy artificial intelligence software at Partners' largest hospitals: Massachusetts General Hospital and Brigham and Women's Hospital. "This is about creating digital tools that will have a profound impact on medicine, said John Flannery, chief executive of GE Healthcare. "By leveraging AI across every patient interaction, workflow challenge and administrative need, this collaboration will drive improvements in quality, cost and access." The new technologies could reduce unnecessary procedures such as some biopsies and automate tedious medical image review, the companies said. The first use of the technology will focus on medical images from x-rays, MRIs and other scans, which would help determine the impact of a stroke, quickly identify emergency room patients with fractures and help track how tumors respond to new cancer treatments.

Curricula for teaching machine learning have existed for decades and even more recent technical subjects (deep learning or big data architectures) have almost standard course outlines and linearized storylines. On the other hand, teaching support for the data science process has been elusive, even though the outlines of the process have been around since the 90s. Understanding the process requires not only wide technical background in machine learning but also basic notions of businesses administration. I have elaborated on the organizational difficulties of data science transformation stemming from these complexities in a previous essay; here I will share my experience on teaching the data science process. I recently had the opportunity to try some experimental pedagogical techniques on about hundred top tier engineering students from Ecole Polytechnique.

Google I/O, the company's annual developer conference, is the company's biggest event of the year. One of the main talking points at this year's conference will likely be artificial intelligence. Google has been slowly injecting AI into many of its products and services and the company's CEO Sundar Pichai has sounded very bullish on the prospects of artificial intelligence in recent public appearances. As our chart illustrates, Google's recent M&A activity also speaks its ambitions in the AI field. According to numbers compiled by CB Insights, the company acquired 11 artificial intelligence startups since 2012, more than any other company.

CylancePROTECT will be integral to the new Razberi CameraDefense solution that, combined with Razberi's secure appliance architecture, provides comprehensive protection over the server, video management systems (VMS), and camera ecosystem. "The physical and network security worlds continue to converge, putting video surveillance systems and any attached networks at risk from unprotected endpoints," said Tom Galvin, Razberi CEO. "CylancePROTECT is ideal for the Razberi distributed architecture, enabling us to offer our customers the most advanced system for anti-virus protection." CylancePROTECT leverages artificial intelligence to detect and prevent malware from executing on endpoints in real time. Because it uses very little memory and less than one percent of CPU, CylancePROTECT will not disrupt the video management systems running on Razberi ServerSwitchIQ appliances.